Long-horizon personalization requires dialog assistants to retrieve user-specific facts from extended interaction histories. However, many relevant facts have low semantic similarity to the query under dense retrieval. Standard Retrieval-Augmented Generation (RAG) and GraphRAG systems remain largely retrospective: they rely on embedding similarity or fixed graph traversals, often missing relevant facts that lie far from the query in embedding space. Inspired by prospection, the human ability to use imagined futures as cues for recall, we introduce Prospection-Guided Retrieval (PGR), which decouples retrieval from memory storage. Given a user query, PGR expands the goal into a linear chain or tree of thoughts (TOT) denoting plausible next steps, and uses these steps as retrieval probes instead of the original query alone. Retrieved facts are then used to personalize the next round of prospection, enabling PGR to uncover additional memories by grounding the simulation in the user’s history. We also introduce MemoryQuest, a challenging multi-session benchmark where each query is annotated with 3–5 dated reference facts under a low query-reference similarity constraint. Across 1,625 queries spanning 185 user profiles from 3 public datasets, PGR-TOT substantially improves retrieval, achieving nearly 3x recall on MemoryQuest over baselines. In pairwise LLM-as-judge evaluations, PGR-generated responses are preferred on 89–98% of queries, with blinded human annotations showing a similar trend. Overall, the results show that explicit prospection yields large gains in long-horizon retrieval and response quality over similarity-only baselines.

A Large Language Model (LLM) offers versatility across domains and tasks, purportedly benefiting users with a wide variety of behaviors and preferences. We question this perception about an LLM when users have inherently subjective behaviors and preferences, as seen in their ubiquitous and idiosyncratic browsing of websites or apps. The sequential behavior logs of pages, thus generated, form something akin to each user’s self-constructed “language”, albeit without the structure and grammar imbued in natural languages. We ask: (i) Can a small LM represent the “language of browsing” better than a large LM? (ii) Can an LM with a single set of parameters (or, single LM) adequately capture myriad users’ heterogeneous, subjective behaviors and preferences? (iii) Can a single LM with high average performance, yield low variance in performance to make alignment good at user level? We introduce clusterwise LM training, HeTLM (Heterogeneity aware Training of Language Model), appropriate for subjective behaviors. We find that (i) a small LM trained using a page-level tokenizer outperforms large pretrained or finetuned LMs; (ii) HeTLM with heterogeneous cluster specific set of parameters outperforms a single LM of the same family, controlling for the number of parameters; and (iii) a higher mean and a lower variance in generation ensues, implying improved alignment.